Remote management method and system based on video image processing

Abstract

The invention discloses a remote management method and system based on video image processing, and the method comprises the steps: constructing a disease course dynamic tracking tree based on a disease course template tree and interaction information; in the disease course dynamic tracking tree, nodes are used for describing sub-links of treatment, the sub-links are image data or stage diagnosis results, and edges are used for describing relevance among the sub-links; determining black, white and grey attributes of each node based on the preliminary identification result, and determining a tracing path of each node based on the black, white and grey attributes of each node in combination with a depth-first search algorithm; wherein the gray nodes are stored in the cache module; black nodes are stored in a relevance database in a manner referring to integrity constraints. According to the method, a depth-first search algorithm is creatively combined with a disease course dynamic tracking tree, so that the problems of expert identification delay and quick control of stored data are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of scalp hair loss treatment, and particularly relates to a remote management method and system based on video image processing. Background Art

[0002] With the continuous popularization of telemedicine and the development of 5G networks and high-definition camera devices, remote management systems based on video image processing have increasingly become an important supplementary means of medical services. This system uses camera devices to collect scalp videos of patients, and through real-time image processing and data transmission, it realizes the rapid diagnosis of hair loss conditions by doctors and the dynamic adjustment of treatment plans. The system can not only break through geographical restrictions and achieve cross-regional sharing of medical resources, but also use algorithms such as image enhancement and feature extraction to finely analyze the scalp state, providing personalized and whole-process monitored treatment services for patients. In the past, traditional treatments relied on in-person visits and it was difficult to achieve continuous monitoring. The popularization and application of this remote scalp hair loss treatment management system is expected to significantly improve the diagnostic accuracy and treatment efficiency, providing strong support for the in-depth expansion of telemedicine in the fields of skin diseases, hair loss, etc.

[0003] Currently, remote scalp hair loss management mainly uses webcams to collect video images and improves the image quality through preprocessing (such as noise filtering, histogram equalization) (the proportion of unclear image data problems is about 10%). For example, the commonly used adaptive threshold algorithm (setting the threshold T to dynamically adjust within the range of 50-150) is used to process images, but it is still affected by environmental lighting and different devices, resulting in the loss of image details. More prominent is the data backtracking problem. During the treatment monitoring process, doctors need to rely on historical video data to compare and judge the progress of patients' hair loss. However, currently, the data storage standards of each remote platform are not unified, and the collected time-series data lacks a unified index, resulting in data breaks or discontinuous timestamps, making it difficult to completely track the evolution of the patient's condition. Some solutions attempt to use time series models (such as tracking algorithms based on state space models) to perform compensation calculations on the data, but due to the lack of information between video frames and algorithm calculation errors, it is difficult to meet the requirements of accurate backtracking. Overall, although there are certain data preprocessing and backtracking mechanisms, a mature and reliable closed-loop management system for the storage, standardization, and time-series analysis of historical image data has not yet been formed, which directly affects doctors' continuous tracking and evaluation of the treatment process of patients. Summary of the Invention

[0004] To solve the deficiencies in the prior art, the purpose of the present invention is to solve the above-mentioned defects and further propose a remote management method and system based on video image processing.

[0005] The present invention adopts the following technical solutions.

[0006] The first aspect of the present invention discloses a remote management method based on video image processing, including: constructing a disease course dynamic tracking tree based on a disease course template tree and interaction information; in the disease course dynamic tracking tree: nodes are used to describe sub-links of treatment, and the sub-links are image data or phased diagnosis results, and edges are used to describe the relevance between sub-links; determining the black, white, and gray attributes of each node based on the preliminary identification result, and determining the trace path of each node by combining the depth-first search algorithm; among them, gray nodes are stored in a cache module; black nodes are stored in a relational database in a way of referential integrity constraint.

[0007] Specifically, gray nodes are not stored in the relational database in a way of referential integrity constraint.

[0008] Specifically, if the final identification result of a gray node is consistent with the preliminary identification result, set the gray node to black and clear the gray node from the cache module.

[0009] Specifically, if the first node corresponds to image data and the preliminary identification result is not credible, generate and send a second node; if the preliminary identification result of the second node is credible, store the second node in the relational database in a way of referential integrity constraint; based on the depth-first search algorithm, trace the gray node and clear the cache corresponding to the gray node.

[0010] Specifically, if the sub-link is image data, the preliminary identification result is determined based on the ResNet-50 network, which specifically includes: image preprocessing and LED parameter extraction; based on the preprocessed image, deep image features are extracted by the ResNet-50 network; the deep image features are fused with the LED parameters, and classification is performed through the fully connected layer of the ResNet-50 network to obtain the final detection result; if all indicators of the detection result are greater than a preset confidence threshold, it is determined that the preliminary identification result is credible.

[0011] Specifically, if the gray node corresponds to a phased diagnosis result and the final identification result is inconsistent with the preliminary identification result, set the gray node to black and clear the gray node from the cache module; use the final identification result as the pathological classification standard and whether the two identification results are consistent as the sample data classification standard, and store the cache and index of the gray node in the pathological classification table and the sample classification table in the relational database in a way of referential integrity constraint respectively.

[0012] Specifically, using the phased diagnosis result as the main node of the disease course dynamic tracking tree and the image data as the image data, specifically includes: when the main node corresponding to the phased diagnosis result is set to a black node, use the image data associated with the phased diagnosis result as the slave node mounted on the main node.

[0013] The second aspect of the present invention discloses a remote management system based on video image processing for implementing the method described in the first aspect. The system includes a plurality of remote terminals and a plurality of client terminals; among them, the remote terminals adopt a star-shaped hierarchical structure and include sub-remote terminals. The remote terminal is used to construct a dynamic disease course tracking tree based on the disease course template tree and interaction information; in the dynamic disease course tracking tree: the nodes are used to describe the sub-links of the treatment, the sub-links are image data or phased diagnosis results, and the edges are used to describe the relevance between the sub-links. In response to the sub-nodes generated by the client terminal, the sub-remote terminal determines the black, white, and gray attributes of each node based on the preliminary identification result. Based on the black, white, and gray attributes of each node and in combination with the depth-first search algorithm, the tracing path of each node is determined; among them, the sub-remote terminal stores the gray nodes in the cache module; the remote terminal stores the black nodes in the correlation database in a way of referential integrity constraint.

[0014] The third aspect of the present invention discloses a terminal, including a processor and a storage medium; it is characterized in that: The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the method described in the first aspect.

[0015] The fourth aspect of the present invention discloses a computer-readable storage medium, on which a computer program is stored. It is characterized in that when the program is executed by a processor, the steps of the method described in the first aspect are realized.

[0016] The beneficial effects of the present invention: Compared with the prior art, the present invention first discloses a remote management concept based on video image processing. Different from the traditional disease course tracking chain, the present invention first forms a preliminary identification result based on an algorithm. If the standard confidence level is not reached, then the final identification result is given by experts and professors. For the delay in this process, the depth-first search algorithm is creatively used in combination with the dynamic disease course tracking tree to solve it. On this basis, in the application scenario, the image data is obtained by a multi-spectral LED light source array, and the nodes of the tree are also image data or phased diagnosis results with integrity constraint relationships. It should be noted that the dynamic disease course tracking tree in the present invention is not only used for dynamic disease course tracking, but actually undertakes the mapping function of the nodes. In addition, the disease course tracking tree integrates image data and phased diagnosis results, making it easier for medical staff to analyze the patient's process; furthermore, the present invention creatively adopts the idea of the depth-first search algorithm and grafts the mapping idea onto the dynamic disease course tracking tree, so that no additional cost information (such as a mapping table) needs to be added. Whether it is to backtrack and clear incorrect image data or re-classify defective sample data, fast manipulation of the stored data can be achieved. Description of the Drawings

[0017] Figure 1A is a schematic flowchart of a traditional treatment process.

[0018] Figure 1B is a schematic flowchart of the treatment process of an embodiment of the present invention.

[0019] Figure 2A is a schematic diagram of a branch node of the disease course dynamic tracking tree of an embodiment of the present invention.

[0020] Figure 2B is another schematic diagram of the disease course dynamic tracking tree of an embodiment of the present invention.

[0021] Figure 3 is a schematic diagram of a remote management system based on video image processing according to an embodiment of the present invention. Detailed Embodiments

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] The remote management system involved in the present invention is used for the treatment of scalp hair loss. In the prior art, a remote management system based on video image processing, as shown in FIG. 1, may include: a plurality of remote terminals and a plurality of client terminals. The remote terminal should at least include a multimedia data storage module and a clinical process monitoring module. The multimedia data storage module is used to record and store image data; the clinical process monitoring module is used to track and record the entire onset and treatment process of a single patient in real time, forming a disease course tracking chain for remote analysis and guidance of the overall case.

[0024] In the usage scenario of the present invention, the remote terminal adopts a star - shaped hierarchical structure, that is, coordinated by a central remote terminal, and the remaining remote terminals participate in data exchange as slave nodes. The specific network topology structure is determined according to the actual deployment environment and will not be elaborated further. It can be understood that due to the star - shaped hierarchical structure of the remote terminal, the data consistency of the disease course tracking chain needs to be ensured in each sub - remote terminal.

[0025] The client terminal can be a mobile phone APP or a multi - spectral light source acquisition terminal. Different from the mobile phone APP, the multi - spectral light source acquisition terminal includes: a multi - spectral LED light source array and a communication module.

[0026] The multi-spectral LED light source array may include: white light LED, red light LED, blue light LED, etc.; the multi-spectral LED light source array adopts a matrix arrangement. Each LED unit can be independently controlled to ensure the optimization of imaging details in different bands. LED parameters may include: light intensity value, PWM frequency, duty cycle, etc.; among them, the LED sets the light intensity value according to the specific detection requirements, for example: white light LED adopts D65 standard (daylight standard) to simulate natural light at 6500K color temperature, and the brightness is controlled at 500-1500 lm / m²; ultraviolet LED (365~405nm) should strictly control its output power to ensure radiation safety. The current industry standard recommends that ultraviolet radiation should not exceed 15 µW / cm² to ensure human safety.

[0027] White light LEDs simulate standard daylight to ensure the overall brightness and color restoration of the image, making it easier for subsequent algorithms to distinguish the color boundary between the scalp and the hair. Red light LEDs (620-750nm) are used to compare the bleeding phenomenon between the scalp and the base of the hair. Blue light LEDs (450-495nm) can enhance surface details and are suitable for distinguishing between exposed scalp parts and covered hair. Green light LEDs (500-570nm) are extremely sensitive to the fiber structure, color and diameter of hair and are used to accurately detect the boundary between hair and scalp and the density of hair. Ultraviolet LEDs (365-405nm) are used to enhance the imaging of the hair surface, cuticle and fine cracks (such as hair scales) under low power conditions. Near-infrared LEDs are used for monitoring hair follicles and microcirculation under the scalp. Near-infrared light has a strong penetrating ability and is sensitive to the spectral response of the internal tissue of the hair. In an embodiment of the present invention, a dual-band near-infrared design can be used: (1) short-wave band (750-850nm) (1) Long-wavelength band (850–950 nm): suitable for detecting subscalp structures; (2) Long-wavelength band (850–950 nm): for penetrating imaging of the interior of hair fibers and distinguishing the connection between the hair shaft and the hair follicle.

[0028] The communication module supports wired or wireless transmission protocols and can transmit the collected video or image data to the remote terminal in real time to achieve remote monitoring and instant diagnosis.

[0029] Traditional treatment procedures, such as Figure 1A As shown, an exemplary method may include steps 1 to 4.

[0030] Step 1, the client generates and sends first image data as a first node; The image data may include the first image data and the second image data described below. The image data is usually the patient's scalp video information; the image data may be a single picture or a picture sequence or a video, which may record the initial state of the scalp and also include the change information before and after medication during the treatment process.

[0031] Step 2, store the first image data in the relational database in the manner of referential integrity.

[0032] It should be understood that the treatment process of scalp hair loss itself is not complex. The complexity lies in forming high-quality image data and ensuring the interactivity between the remote terminal and the client.

[0033] Among them, the referential integrity constraints include: the constraints between the nodes in the disease course tracking chain (if the node corresponds to image data) and the image data, and the constraints between the nodes in the disease course tracking chain (if the node corresponds to the stage diagnosis result) and the pathological classification.

[0034] In the specific practice process, the remote terminal is actually in the form of a visual interface for medical staff to conduct analysis. When opening the window of a certain patient, first give the disease course tracking chain corresponding to the patient, and then for the target node, usually the last node represents the most real-time state. For the patient's question information, or directly analyze the image data to give guiding feedback information. Taking the constraints between the nodes in the disease course tracking chain and the image data as an example, in the interface, medical staff actually obtain the corresponding image data or interaction information by clicking on the target node or target edge. This means that between the nodes in the disease course tracking chain and the image data, they must conform to the referential integrity constraints. It is not difficult to analogize that the sample data corresponding to the pathological classification (essentially part or all of the data of the disease course tracking chain) should also be mapped to the real nodes to facilitate medical staff to statistically analyze patients with similar diseases. That is, between the nodes of the disease course tracking chain and the pathological classification, they must also conform to the referential integrity constraints. That is to say, each node formed by the patient should be a whole. When medical staff analyze a certain node (the corresponding data information), they should be able to obtain all the data information of the patient based on the disease course tracking chain, and then analyze from a global perspective. That is, the node and its corresponding data information must be constrained.

[0035] Taking the constraints between the nodes in the disease course tracking chain and the image data as an example, the code of the referential integrity constraint can be as follows: CREATE TABLE MediaInfo( NodeID INT, MediaInfomation BLOB ); CREATE TABLE Node ( NodeID INT, FaNodeID INT, ChildNodeID STRING, CONSTRAINT FOREIGN KEY (NodeID) REFERENCES MediaInfo ON DELETE CASCADE ON UPDATE CASCADE ); Among them, the nodes in the disease course tracking chain represent NodeID, and the image data is MediaInfomation.

[0036] It is not difficult to understand that the disease course tracking chain, image data, and interaction information cannot be stored in the same table in the database. This is because the image data may be very large, or even video data. If they are mixed together, it does not conform to the design principles of the table. In addition, it is difficult to guarantee the refresh process and consistency constraints of the disease course tracking chain. Among them, the interaction information may include the guidance feedback information in the following text, as well as the patient's question information, etc.

[0037] Step 3, generate guidance feedback information based on the first node.

[0038] The guidance feedback information can be the result generated by the scalp hair loss treatment management system based on the algorithm processing of video images, or the guidance opinions given by professional medical staff after watching the scalp video information. The guidance feedback information can be a suggestion to retake supplementary shots of different frequency bands of LED for local areas to obtain more accurate original data. For example: for alopecia areata cases, it can be recommended to locally increase the blue light output (the blue light brightness is adjusted to about 350 lm / m²) to enhance the local texture of the scalp and hair follicle details; for diffuse alopecia cases, it is recommended to appropriately reduce the ultraviolet LED output and enhance the red LED (the red light output is adjusted to a medium brightness level) to avoid local high reflection interference. In addition, the specific parameter settings of each LED should also be described in detail in the feedback information. For example: for areas with sparse hair, it is recommended to increase the green LED light intensity by 10%-20% to more clearly collect hair details.

[0039] Step 4, generate second image data based on the guidance feedback information as the second node.

[0040] It can be understood that in some embodiments, the second image data can be used as the first image data, so as to trace back to Step 1 and iterate again to form more accurate scalp original information. The motivation for forming the iteration can be either to enhance the effect of the original first image data or to directly replace the invalid first image data. This iteration mechanism can not only enhance the initial imaging effect but also effectively eliminate the image data that is poorly collected due to environmental interference.

[0041] Since there may be a long delay between the final identification result and the preliminary identification result, it is actually quite difficult to remove the first image data subsequent to step 4. Understandably, the first image data may have more than one node, most likely including multiple parallel nodes, or multiple serial nodes, or a combination of both.

[0042] Based on this, the present invention discloses a remote management method based on video image processing, including: constructing a disease course dynamic tracking tree based on a disease course template tree and interaction information; in the disease course dynamic tracking tree: nodes are used to describe sub-links of treatment, and the sub-links are image data or stage diagnosis results, and edges are used to describe the relevance between sub-links; determining the black-white-gray attributes of each node based on the preliminary identification result, and based on the black-white-gray attributes of each node, combining with the depth-first search algorithm to determine the traceability path of each node; among them, the image data corresponding to the gray nodes is stored in the cache module; the image data of the black nodes is stored in the relational database in a way of referential integrity constraint.

[0043] It should be noted that the significance of determining the gray nodes in the above is to facilitate tracing using the depth-first search algorithm. Since the disease course dynamic tracking tree is usually stored in a cache system (Etcd) with high consistency requirements, therefore, on the one hand, a mapping table will not be established in practice to associate the traceability path of the disease course dynamic tracking tree; on the other hand, a cache system with high consistency requirements usually does not support this storage method similar to integrity constraint either. It should be understood that the present invention uses the depth-first search algorithm not to find the shortest path, but to obtain its traceability path under the premise of determining the black, white, and gray nodes, that is, in the depth-first search algorithm, the path formed during the (second) traversal is returned. Since the specific process of the depth-first search algorithm is not elaborated in the present invention.

[0044] Understandably, the gray nodes are not stored in the relational database in a way of referential integrity constraint.

[0045] In some embodiments, the cache module can be either redis or the relational database itself, but a new temporary table needs to be created to distinguish the storage in the relational database in a way of referential integrity constraint.

[0046] Understandably, the gray nodes represent sub-links where the preliminary identification result is not credible and treatment has been received, the black nodes represent sub-links where the preliminary identification result is credible and treatment has been received; the white nodes represent sub-links where treatment has not been received.

[0047] For a scalp hair loss treatment management system, the preliminary identification result is usually given by medical staff assisted by an intelligent algorithm, and the final identification result can be given by expert professors. Both the preliminary identification result and the final identification result correspond to a single node.

[0048] For image data, an image quality detection algorithm based on a convolutional neural network can be adopted. For example, the preliminary identification result can be determined based on the ResNet-50 network. The specific process is as follows: image preprocessing and LED parameter extraction; based on the preprocessed image, deep image features are extracted by the ResNet-50 network; the deep image features are fused with the LED parameters, and classification is performed through the fully connected layer of the ResNet-50 network to obtain the final detection result. It can be understood that if each index of the detection result is greater than the preset confidence threshold, the preliminary identification result is determined to be credible. Since the ResNet-50 network is a known algorithm, the specific process will not be elaborated here.

[0049] For the stage diagnosis result, the preliminary identification result is determined based on the XGBoost algorithm, which specifically includes: performing data preprocessing on the positive and negative sample data in the sample classification table, and extracting features to obtain a mapping vector; using the cross-entropy loss function to train the mapping vector, and optimizing the parameters of the XGBoost algorithm based on the k-fold cross-validation method. Since the XGBoost algorithm is a known algorithm, the specific process will not be elaborated here.

[0050] In some embodiments, the patient's ID and the disease course dynamic tracking tree can be stored in the key-value storage system as key-value pairs respectively. In some embodiments, the key-value storage system can be Etcd. Etcd is a distributed reliable key-value storage system written in the Go language, which is often used to store data that needs to be highly available and consistent in a distributed system. Etcd can be understood as a distributed notepad.

[0051] The disease course template tree refers to the general initial diagnosis process given according to the preliminary diagnosis information, and the concrete manifestation of the initial diagnosis process is the initial disease course dynamic tracking tree. It can be understood that the disease course dynamic tracking tree is not pre-set, but changes in real time according to the interaction information. For example, during the treatment process, when there is a treatment plan adjustment or complication (such as local inflammation deterioration) in the patient, the system can automatically form branches in the disease course record, recording the initial treatment plan on the one hand and the adjusted plan on the other hand; the branching situation can be recorded in the way of a linked list for continuous events or in a tree-like record structure, where: the linked list record is suitable for cases with a single continuous treatment process, recording each image data and feedback information; the tree-like record structure is suitable for cases with multiple plan disagreements during treatment. For example, when the condition recurs or new pathological changes occur during the treatment process, different sub-paths are branched out. In addition, for some difficult cases, researchers can mark the corresponding tags at the corresponding records for later special research on the case, such as analyzing the recurrence reason or testing new treatment plans.

[0052] Such as Figure 2AAs shown, the node is composed of a sub-link of the client's treatment. The sub-link here is not limited to a specific scope, that is, the node itself can include a main node and a sub-node, where the main node is composed of multiple sub-nodes. In other words, since the dynamic tracking tree of the disease course changes in real time, in some specific examples, the initial dynamic tracking tree of the disease course can be described as {a->b; b->c}, and in the subsequent diagnosis and treatment, a new disease is found at the b node or the a node, which may cause differences in the treatment plan, and the subsequent dynamic tracking tree of the disease course will eventually split into {a->b1; b1->b2; b1->b3; b2->c; b3->d}. Among them, b is the main node, and b1 and b2 are sub-nodes.

[0053] The edges in the disease course dynamic tracking tree mainly refer to guidance feedback information, which determines the association between the current node (eg, the first node) and the next node to be generated (eg, the second node).

[0054] In some embodiments, if the final identification result of the gray node is consistent with the preliminary identification result, the gray node is set to black, and the data corresponding to the gray node in the cache module is cleared at the same time, and transferred to the associative database in a referential integrity constraint manner.

[0055] In other embodiments, if the first node corresponds to image data and the preliminary identification result is unreliable, the logic here can refer to Figure 1B As shown, in the disease course dynamic tracking tree, the sub-path is re-branched (for example: if the first node is Figure 2A b2, then the second node is Figure 2A b3), that is, generating and sending the second node; if the preliminary identification result of the second node is credible, the image data or phased diagnosis result corresponding to the second node is stored in the associative database in the form of reference integrity constraints; based on the depth-first search algorithm, the gray node, that is, the first node, is traced back, the cache corresponding to the gray node is cleared, and the gray node is cleared. It can be understood that if the preliminary identification result of the second node is still unreliable, the backtracking iteration is continued: the subpath is re-branched to generate and send the third node.

[0056] In some other embodiments, if the grey node corresponds to a phased diagnosis result and the final identification result is inconsistent with the preliminary identification result, the grey node is still set to black, and the grey node is cleared from the cache module; taking the final identification result as the pathological classification standard and whether the two identification results are consistent as the sample data classification standard, the cache and index of the grey node (which is set to black) are stored in the pathological classification table and the sample classification table in the relational database in the manner of referential integrity constraints. It can be understood that the disease course dynamic tracking tree corresponding to each patient should contribute to the establishment of the algorithm model. Therefore, it is necessary to perform pathological classification on the newly generated disease course dynamic tracking tree as the basis for subsequent algorithm update and iteration. In the medical field, the algorithm model should at least be able to perform feedback learning. That is to say, the algorithm model needs to learn and analyze negative sample data. Even if the new algorithm model cannot make an accurate judgment, it should also perform special discrimination on negative sample data. For example: it is impossible to obtain the corresponding conclusion based on the negative sample model, which can greatly reduce the misjudgment rate (it is better not to be able to judge than to reduce misjudgment as much as possible). Therefore, in the present invention, pathological classification actually mainly refers to whether the sample data is negative sample data. It can be understood that negative sample data is the sample data for which the algorithm classification fails (the preliminary identification result is inconsistent with the final identification result).

[0057] Note: The cache of the grey node can refer to the data stored by the grey node in the cache system, such as image data or phased diagnosis results. Since the data in the cache system cannot be saved twice in the relational database, the above-mentioned index of the grey node should be understood as the index of the cache of the grey node.

[0058] In some more preferred embodiments, the disease course dynamic tracking tree can be as Figure 2B shown, taking the phased diagnosis result as the main node of the disease course dynamic tracking tree and the image data as the subordinate node of the disease course dynamic tracking tree; specifically, considering that when the expert professor gives the phased diagnosis result, the final identification result of the associated image data is also necessarily given at the same time. Therefore, the phased diagnosis result can be directly used as the main node of the disease course dynamic tracking tree. And when the main node corresponding to the phased diagnosis result is set to black, the image data associated with the phased diagnosis result is used as the subordinate node (for example: Figure 2B in m2) mounted on this main node (for example: Figure 2B in c4).

[0059] It can be understood that in this embodiment, at any moment, the grey subordinate node must be the subordinate node that is not mounted on the main node (for example: Figure 2BAmong c1 to c3), at this time, m2 has not yet formed a final identification result. Therefore, on the one hand, this approach can simplify the dynamic tracking logic of the disease course dynamic tracking tree without the need for additional marking constraints; on the other hand, it also reduces the height of the tree, making the dynamic tracking process of the disease course dynamic tracking tree essentially composed only of the nodes corresponding to the phased diagnosis results, thereby greatly reducing the burden on Etcd. Among them, c4 and c2 are the same node, and are only described separately for convenience of explanation. In Figure 2B it is assumed that c2 finally changes from a gray node to a black node, and c1 and c3 do not change to black nodes; therefore, c2 is retained, that is, c4, while c1 and c3 are discarded. The image data associated with the phased diagnosis results (such as c1 to c3) generally refers to the image data temporally (i.e., in the disease course dynamic tracking tree) between the previous phased diagnosis result (such as m1) and this phased diagnosis result (such as m2).

[0060] Correspondingly, the present invention also discloses a remote management system based on video image processing, as Figure 3 shown, including a plurality of remote terminals and a plurality of clients; among them, the remote terminals adopt a star hierarchical structure and include sub-remote terminals; The remote terminal is used to construct a disease course dynamic tracking tree based on the disease course template tree and interaction information; in the disease course dynamic tracking tree: the nodes are used to describe the sub-links of the treatment, the sub-links are image data or phased diagnosis results, and the edges are used to describe the relevance between the sub-links; In response to the sub-nodes generated by the client, the sub-remote terminal determines the black, white, and gray attributes of each node based on the preliminary identification result, and based on the black, white, and gray attributes of each node, combines the depth-first search algorithm to determine the trace path of each node; among them, the sub-remote terminal stores the gray nodes (such as: Figure 3 g1 in) in the cache module; the remote terminal stores the black nodes (such as: Figure 3 b1 in) in the association database in a way of referential integrity constraint.

[0061] In the above embodiments, the descriptions of each embodiment have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combinations of these technical features do not conflict, they should all be considered as the scope recorded in this specification.

[0062] Those skilled in the art will readily conceive of other embodiments of the present application after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application, which follow the general principles of the present application and include well-known common general knowledge or conventional technical means in the technical field not disclosed in the present application. The specification and examples are only regarded as exemplary, and the true scope and spirit of the present application are pointed out by the following claims.

[0063] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A remote management method based on video image processing, characterized in that: include: Based on the disease course template tree and interactive information, a dynamic disease course tracking tree is constructed; In the dynamic tracking tree of the disease course: nodes are used to describe the sub-links of treatment, which are image data or stage-by-stage diagnostic results, and edges are used to describe the associations between sub-links; the black, white and gray attributes of each node are determined based on the preliminary identification results, and the traceability path of each node is determined based on the black, white and gray attributes of each node combined with the depth-first search algorithm; among them, gray nodes are stored in the cache module; black nodes are stored in the associative database in the form of reference integrity constraints.

2. A remote management method based on video image processing according to claim 1, characterized in that: Gray nodes are not stored in the relational database with referential integrity constraints.

3. The remote management method based on video image processing according to claim 1, characterized in that: If the final identification result of the gray node is consistent with the preliminary identification result, the gray node is set to black and the gray node is cleared from the cache module.

4. The remote management method based on video image processing according to claim 1, characterized in that: If the first node corresponds to the image data and the preliminary identification result is not credible, generating and sending a second node; If the preliminary identification result of the second node is credible, the second node is stored in the relational database in a referential integrity constraint manner; Based on the depth-first search algorithm, trace back the gray nodes and clear the cache corresponding to the gray nodes.

5. The remote management method based on video image processing according to claim 1, characterized in that: If the sub-link is image data, the preliminary identification result is determined based on the ResNet-50 network, specifically including: image preprocessing and LED parameter extraction; based on the preprocessed image, the ResNet-50 network extracts deep image features; the deep image features are fused with the LED parameters, and classified through the fully connected layer of the ResNet-50 network to obtain the final detection result; if all indicators of the detection result are greater than the preset confidence threshold, the preliminary identification result is judged to be credible.

6. The remote management method based on video image processing according to claim 1, characterized in that: If the gray node corresponds to the staged diagnosis result, and the final identification result is inconsistent with the preliminary identification result, the gray node is set to black and the gray node is cleared from the cache module; The final identification result is used as the pathological classification standard, and the consistency of the two identification results is used as the sample data classification standard. The cache and index of the gray nodes are stored in the pathological classification table and the sample classification table in the relational database in the form of referential integrity constraints.

7. The remote management method based on video image processing according to claim 1, characterized in that: The stage diagnosis result is used as the main node of the dynamic tracking tree of the disease course, and the image data is used as the image data. Specifically, when the main node corresponding to the stage diagnosis result is set as a black node, the image data associated with the stage diagnosis result is used as a slave node mounted on the main node.

8. A remote management system based on video image processing, used to execute the method according to any one of claims 1 to 7, characterized in that: The system includes a plurality of remote terminals and a plurality of clients; wherein the remote terminal adopts a star-shaped hierarchical structure, including sub-remote terminals; The remote terminal is used to construct a dynamic disease tracking tree based on the disease template tree and the interactive information; in the dynamic disease tracking tree: nodes are used to describe sub-links of treatment, which are image data or staged diagnosis results, and edges are used to describe the associations between sub-links; In response to the child nodes generated by the client, the child remote terminal determines the black, white and gray attributes of each node based on the preliminary identification results, and based on the black, white and gray attributes of each node, combined with the depth-first search algorithm, determines the traceability path of each node; wherein, the child remote terminal stores the gray nodes in the cache module; the remote terminal stores the black nodes in the associative database in a reference integrity constraint manner.

9. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images